Design and Validation of a Modified Delayed Output Feedback Controller for Hip Exoskeleton Assistance

This study proposes a control algorithm, Modified Delayed Output Feedback Control (MDOFC), for energy-efficient hip assistance without explicit gait phase estimation. Conventional gait-assistive controllers often use pre-defined torque profiles synchronized with estimated gait phase, but these are vulnerable to changes such as directional turns. The Delayed Output Feedback Control (DOFC) addressed this issue by generating phase-free assistive torque using delayed and amplified joint angles; however, it could only provide torque profiles of a simple form. MDOFC overcomes this by adding harmonic components to produce more adaptive torque profiles. Parameters were optimized via forward dynamic gait simulation with a reinforcement learning-based neural network controller. Simulations showed MDOFC increased net metabolic cost reduction using only two parameters, similar to DOFC, and maintained dynamic stability up to 45 W based on the short-term Lyapunov exponent. Human experiments confirmed comparable metabolic cost reduction to DOFC, with 8.8% less mechanical power consumption. These results indicate MDOFC as an effective control framework for wearable hip exoskeletons.